feat: 除 PNCA 解码外全上 NPU — 新增 pitch_energy/duration/postnet 模型,host_weights 精简到 215
Browse files- .gitattributes +3 -0
- model/duration.axmodel +3 -0
- model/host_weights/manifest.json +16 -366
- model/pitch_energy.axmodel +3 -0
- model/postnet.axmodel +3 -0
.gitattributes
CHANGED
|
@@ -36,3 +36,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 36 |
model/am_enc.axmodel filter=lfs diff=lfs merge=lfs -text
|
| 37 |
model/voc.axmodel filter=lfs diff=lfs merge=lfs -text
|
| 38 |
sdk/axrt/lib/libax_engine.so filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
model/am_enc.axmodel filter=lfs diff=lfs merge=lfs -text
|
| 37 |
model/voc.axmodel filter=lfs diff=lfs merge=lfs -text
|
| 38 |
sdk/axrt/lib/libax_engine.so filter=lfs diff=lfs merge=lfs -text
|
| 39 |
+
model/duration.axmodel filter=lfs diff=lfs merge=lfs -text
|
| 40 |
+
model/pitch_energy.axmodel filter=lfs diff=lfs merge=lfs -text
|
| 41 |
+
model/postnet.axmodel filter=lfs diff=lfs merge=lfs -text
|
model/duration.axmodel
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:419d5bccf01f5b5d771c7e10610051ca1c835f6f776939c1b6f8308d4ce52259
|
| 3 |
+
size 1358724
|
model/host_weights/manifest.json
CHANGED
|
@@ -1,159 +1,4 @@
|
|
| 1 |
{
|
| 2 |
-
"dur_fc_b": [
|
| 3 |
-
1
|
| 4 |
-
],
|
| 5 |
-
"dur_fc_w": [
|
| 6 |
-
1,
|
| 7 |
-
128
|
| 8 |
-
],
|
| 9 |
-
"dur_lstm_b_hh0": [
|
| 10 |
-
512
|
| 11 |
-
],
|
| 12 |
-
"dur_lstm_b_hh1": [
|
| 13 |
-
512
|
| 14 |
-
],
|
| 15 |
-
"dur_lstm_b_ih0": [
|
| 16 |
-
512
|
| 17 |
-
],
|
| 18 |
-
"dur_lstm_b_ih1": [
|
| 19 |
-
512
|
| 20 |
-
],
|
| 21 |
-
"dur_lstm_w_hh0": [
|
| 22 |
-
512,
|
| 23 |
-
128
|
| 24 |
-
],
|
| 25 |
-
"dur_lstm_w_hh1": [
|
| 26 |
-
512,
|
| 27 |
-
128
|
| 28 |
-
],
|
| 29 |
-
"dur_lstm_w_ih0": [
|
| 30 |
-
512,
|
| 31 |
-
224
|
| 32 |
-
],
|
| 33 |
-
"dur_lstm_w_ih1": [
|
| 34 |
-
512,
|
| 35 |
-
128
|
| 36 |
-
],
|
| 37 |
-
"dur_pre0_b": [
|
| 38 |
-
128
|
| 39 |
-
],
|
| 40 |
-
"dur_pre0_w": [
|
| 41 |
-
128,
|
| 42 |
-
1
|
| 43 |
-
],
|
| 44 |
-
"dur_pre1_b": [
|
| 45 |
-
128
|
| 46 |
-
],
|
| 47 |
-
"dur_pre1_w": [
|
| 48 |
-
128,
|
| 49 |
-
128
|
| 50 |
-
],
|
| 51 |
-
"energy_blstm_b_hh": [
|
| 52 |
-
512
|
| 53 |
-
],
|
| 54 |
-
"energy_blstm_b_hh_r": [
|
| 55 |
-
512
|
| 56 |
-
],
|
| 57 |
-
"energy_blstm_b_ih": [
|
| 58 |
-
512
|
| 59 |
-
],
|
| 60 |
-
"energy_blstm_b_ih_r": [
|
| 61 |
-
512
|
| 62 |
-
],
|
| 63 |
-
"energy_blstm_w_hh": [
|
| 64 |
-
512,
|
| 65 |
-
128
|
| 66 |
-
],
|
| 67 |
-
"energy_blstm_w_hh_r": [
|
| 68 |
-
512,
|
| 69 |
-
128
|
| 70 |
-
],
|
| 71 |
-
"energy_blstm_w_ih": [
|
| 72 |
-
512,
|
| 73 |
-
128
|
| 74 |
-
],
|
| 75 |
-
"energy_blstm_w_ih_r": [
|
| 76 |
-
512,
|
| 77 |
-
128
|
| 78 |
-
],
|
| 79 |
-
"energy_emb_b": [
|
| 80 |
-
32
|
| 81 |
-
],
|
| 82 |
-
"energy_emb_w": [
|
| 83 |
-
32,
|
| 84 |
-
1,
|
| 85 |
-
9
|
| 86 |
-
],
|
| 87 |
-
"energy_fc_b": [
|
| 88 |
-
1
|
| 89 |
-
],
|
| 90 |
-
"energy_fc_w": [
|
| 91 |
-
1,
|
| 92 |
-
256
|
| 93 |
-
],
|
| 94 |
-
"energy_ffn0_b1": [
|
| 95 |
-
256
|
| 96 |
-
],
|
| 97 |
-
"energy_ffn0_b2": [
|
| 98 |
-
128
|
| 99 |
-
],
|
| 100 |
-
"energy_ffn0_w1": [
|
| 101 |
-
256,
|
| 102 |
-
96,
|
| 103 |
-
1
|
| 104 |
-
],
|
| 105 |
-
"energy_ffn0_w2": [
|
| 106 |
-
128,
|
| 107 |
-
256,
|
| 108 |
-
1
|
| 109 |
-
],
|
| 110 |
-
"energy_ffn1_b1": [
|
| 111 |
-
256
|
| 112 |
-
],
|
| 113 |
-
"energy_ffn1_b2": [
|
| 114 |
-
128
|
| 115 |
-
],
|
| 116 |
-
"energy_ffn1_w1": [
|
| 117 |
-
256,
|
| 118 |
-
128,
|
| 119 |
-
1
|
| 120 |
-
],
|
| 121 |
-
"energy_ffn1_w2": [
|
| 122 |
-
128,
|
| 123 |
-
256,
|
| 124 |
-
1
|
| 125 |
-
],
|
| 126 |
-
"energy_ffn2_b1": [
|
| 127 |
-
256
|
| 128 |
-
],
|
| 129 |
-
"energy_ffn2_b2": [
|
| 130 |
-
128
|
| 131 |
-
],
|
| 132 |
-
"energy_ffn2_w1": [
|
| 133 |
-
256,
|
| 134 |
-
128,
|
| 135 |
-
1
|
| 136 |
-
],
|
| 137 |
-
"energy_ffn2_w2": [
|
| 138 |
-
128,
|
| 139 |
-
256,
|
| 140 |
-
1
|
| 141 |
-
],
|
| 142 |
-
"energy_mem0_conv": [
|
| 143 |
-
128,
|
| 144 |
-
1,
|
| 145 |
-
41
|
| 146 |
-
],
|
| 147 |
-
"energy_mem1_conv": [
|
| 148 |
-
128,
|
| 149 |
-
1,
|
| 150 |
-
41
|
| 151 |
-
],
|
| 152 |
-
"energy_mem2_conv": [
|
| 153 |
-
128,
|
| 154 |
-
1,
|
| 155 |
-
41
|
| 156 |
-
],
|
| 157 |
"l0_fch_b": [
|
| 158 |
128
|
| 159 |
],
|
|
@@ -839,217 +684,6 @@
|
|
| 839 |
240,
|
| 840 |
128
|
| 841 |
],
|
| 842 |
-
"pitch_blstm_b_hh": [
|
| 843 |
-
512
|
| 844 |
-
],
|
| 845 |
-
"pitch_blstm_b_hh_r": [
|
| 846 |
-
512
|
| 847 |
-
],
|
| 848 |
-
"pitch_blstm_b_ih": [
|
| 849 |
-
512
|
| 850 |
-
],
|
| 851 |
-
"pitch_blstm_b_ih_r": [
|
| 852 |
-
512
|
| 853 |
-
],
|
| 854 |
-
"pitch_blstm_w_hh": [
|
| 855 |
-
512,
|
| 856 |
-
128
|
| 857 |
-
],
|
| 858 |
-
"pitch_blstm_w_hh_r": [
|
| 859 |
-
512,
|
| 860 |
-
128
|
| 861 |
-
],
|
| 862 |
-
"pitch_blstm_w_ih": [
|
| 863 |
-
512,
|
| 864 |
-
128
|
| 865 |
-
],
|
| 866 |
-
"pitch_blstm_w_ih_r": [
|
| 867 |
-
512,
|
| 868 |
-
128
|
| 869 |
-
],
|
| 870 |
-
"pitch_emb_b": [
|
| 871 |
-
32
|
| 872 |
-
],
|
| 873 |
-
"pitch_emb_w": [
|
| 874 |
-
32,
|
| 875 |
-
1,
|
| 876 |
-
9
|
| 877 |
-
],
|
| 878 |
-
"pitch_fc_b": [
|
| 879 |
-
1
|
| 880 |
-
],
|
| 881 |
-
"pitch_fc_w": [
|
| 882 |
-
1,
|
| 883 |
-
256
|
| 884 |
-
],
|
| 885 |
-
"pitch_ffn0_b1": [
|
| 886 |
-
256
|
| 887 |
-
],
|
| 888 |
-
"pitch_ffn0_b2": [
|
| 889 |
-
128
|
| 890 |
-
],
|
| 891 |
-
"pitch_ffn0_w1": [
|
| 892 |
-
256,
|
| 893 |
-
96,
|
| 894 |
-
1
|
| 895 |
-
],
|
| 896 |
-
"pitch_ffn0_w2": [
|
| 897 |
-
128,
|
| 898 |
-
256,
|
| 899 |
-
1
|
| 900 |
-
],
|
| 901 |
-
"pitch_ffn1_b1": [
|
| 902 |
-
256
|
| 903 |
-
],
|
| 904 |
-
"pitch_ffn1_b2": [
|
| 905 |
-
128
|
| 906 |
-
],
|
| 907 |
-
"pitch_ffn1_w1": [
|
| 908 |
-
256,
|
| 909 |
-
128,
|
| 910 |
-
1
|
| 911 |
-
],
|
| 912 |
-
"pitch_ffn1_w2": [
|
| 913 |
-
128,
|
| 914 |
-
256,
|
| 915 |
-
1
|
| 916 |
-
],
|
| 917 |
-
"pitch_ffn2_b1": [
|
| 918 |
-
256
|
| 919 |
-
],
|
| 920 |
-
"pitch_ffn2_b2": [
|
| 921 |
-
128
|
| 922 |
-
],
|
| 923 |
-
"pitch_ffn2_w1": [
|
| 924 |
-
256,
|
| 925 |
-
128,
|
| 926 |
-
1
|
| 927 |
-
],
|
| 928 |
-
"pitch_ffn2_w2": [
|
| 929 |
-
128,
|
| 930 |
-
256,
|
| 931 |
-
1
|
| 932 |
-
],
|
| 933 |
-
"pitch_mem0_conv": [
|
| 934 |
-
128,
|
| 935 |
-
1,
|
| 936 |
-
41
|
| 937 |
-
],
|
| 938 |
-
"pitch_mem1_conv": [
|
| 939 |
-
128,
|
| 940 |
-
1,
|
| 941 |
-
41
|
| 942 |
-
],
|
| 943 |
-
"pitch_mem2_conv": [
|
| 944 |
-
128,
|
| 945 |
-
1,
|
| 946 |
-
41
|
| 947 |
-
],
|
| 948 |
-
"post_fc_b": [
|
| 949 |
-
80
|
| 950 |
-
],
|
| 951 |
-
"post_fc_w": [
|
| 952 |
-
80,
|
| 953 |
-
128
|
| 954 |
-
],
|
| 955 |
-
"post_ffn0_b1": [
|
| 956 |
-
512
|
| 957 |
-
],
|
| 958 |
-
"post_ffn0_b2": [
|
| 959 |
-
256
|
| 960 |
-
],
|
| 961 |
-
"post_ffn0_w1": [
|
| 962 |
-
512,
|
| 963 |
-
80,
|
| 964 |
-
1
|
| 965 |
-
],
|
| 966 |
-
"post_ffn0_w2": [
|
| 967 |
-
256,
|
| 968 |
-
512,
|
| 969 |
-
1
|
| 970 |
-
],
|
| 971 |
-
"post_ffn1_b1": [
|
| 972 |
-
512
|
| 973 |
-
],
|
| 974 |
-
"post_ffn1_b2": [
|
| 975 |
-
256
|
| 976 |
-
],
|
| 977 |
-
"post_ffn1_w1": [
|
| 978 |
-
512,
|
| 979 |
-
256,
|
| 980 |
-
1
|
| 981 |
-
],
|
| 982 |
-
"post_ffn1_w2": [
|
| 983 |
-
256,
|
| 984 |
-
512,
|
| 985 |
-
1
|
| 986 |
-
],
|
| 987 |
-
"post_ffn2_b1": [
|
| 988 |
-
512
|
| 989 |
-
],
|
| 990 |
-
"post_ffn2_b2": [
|
| 991 |
-
256
|
| 992 |
-
],
|
| 993 |
-
"post_ffn2_w1": [
|
| 994 |
-
512,
|
| 995 |
-
256,
|
| 996 |
-
1
|
| 997 |
-
],
|
| 998 |
-
"post_ffn2_w2": [
|
| 999 |
-
256,
|
| 1000 |
-
512,
|
| 1001 |
-
1
|
| 1002 |
-
],
|
| 1003 |
-
"post_ffn3_b1": [
|
| 1004 |
-
512
|
| 1005 |
-
],
|
| 1006 |
-
"post_ffn3_b2": [
|
| 1007 |
-
256
|
| 1008 |
-
],
|
| 1009 |
-
"post_ffn3_w1": [
|
| 1010 |
-
512,
|
| 1011 |
-
256,
|
| 1012 |
-
1
|
| 1013 |
-
],
|
| 1014 |
-
"post_ffn3_w2": [
|
| 1015 |
-
256,
|
| 1016 |
-
512,
|
| 1017 |
-
1
|
| 1018 |
-
],
|
| 1019 |
-
"post_lstm_b_hh": [
|
| 1020 |
-
512
|
| 1021 |
-
],
|
| 1022 |
-
"post_lstm_b_ih": [
|
| 1023 |
-
512
|
| 1024 |
-
],
|
| 1025 |
-
"post_lstm_w_hh": [
|
| 1026 |
-
512,
|
| 1027 |
-
128
|
| 1028 |
-
],
|
| 1029 |
-
"post_lstm_w_ih": [
|
| 1030 |
-
512,
|
| 1031 |
-
256
|
| 1032 |
-
],
|
| 1033 |
-
"post_mem0_conv": [
|
| 1034 |
-
256,
|
| 1035 |
-
1,
|
| 1036 |
-
41
|
| 1037 |
-
],
|
| 1038 |
-
"post_mem1_conv": [
|
| 1039 |
-
256,
|
| 1040 |
-
1,
|
| 1041 |
-
41
|
| 1042 |
-
],
|
| 1043 |
-
"post_mem2_conv": [
|
| 1044 |
-
256,
|
| 1045 |
-
1,
|
| 1046 |
-
41
|
| 1047 |
-
],
|
| 1048 |
-
"post_mem3_conv": [
|
| 1049 |
-
256,
|
| 1050 |
-
1,
|
| 1051 |
-
41
|
| 1052 |
-
],
|
| 1053 |
"pre0_b": [
|
| 1054 |
256
|
| 1055 |
],
|
|
@@ -1105,5 +739,21 @@
|
|
| 1105 |
"pos_enc": [
|
| 1106 |
128,
|
| 1107 |
512
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1108 |
]
|
| 1109 |
}
|
|
|
|
| 1 |
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
"l0_fch_b": [
|
| 3 |
128
|
| 4 |
],
|
|
|
|
| 684 |
240,
|
| 685 |
128
|
| 686 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 687 |
"pre0_b": [
|
| 688 |
256
|
| 689 |
],
|
|
|
|
| 739 |
"pos_enc": [
|
| 740 |
128,
|
| 741 |
512
|
| 742 |
+
],
|
| 743 |
+
"pitch_emb_w": [
|
| 744 |
+
32,
|
| 745 |
+
1,
|
| 746 |
+
9
|
| 747 |
+
],
|
| 748 |
+
"pitch_emb_b": [
|
| 749 |
+
32
|
| 750 |
+
],
|
| 751 |
+
"energy_emb_w": [
|
| 752 |
+
32,
|
| 753 |
+
1,
|
| 754 |
+
9
|
| 755 |
+
],
|
| 756 |
+
"energy_emb_b": [
|
| 757 |
+
32
|
| 758 |
]
|
| 759 |
}
|
model/pitch_energy.axmodel
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4e2213b050b725aff33e712390959b6fbfc810ab6d62d1d3f0da0b8011ad36c2
|
| 3 |
+
size 5162287
|
model/postnet.axmodel
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c6d5c094f4c021cc65b39d2fbd6dd8802315c73b72908ebbc2264dd8892121bb
|
| 3 |
+
size 2950487
|